Our paper on ``Federated learning of predictive models from federated Electronic Health Records''...
NeTS-NOSS: SensorNet Architectures for Indoor Location Detection: From Resolution to Robustness
Funding Agency: National Science Foundation, Directorate for Computer and Information Science and Engineering, Division of Computer and Network Systems.
Award Number: CNS-0435312.
Principal Investigators: David Starobinski, Yannis Paschalidis, and Ari Trachtenberg, Boston University.
Technologies for communication, control, and computing have finally converged to the point where it is possible to wirelessly network many tiny sensors to monitor and control a variety of physical systems. Fundamental to such sensor networks (SensorNets) is the ability to accurately and robustly determine the location of sensors and targets. In an outdoor setting, such location detection is often achieved successfully with triangulation or trilateration techniques such as those used in the Global Positioning System (GPS). However, in indoor or dense urban environments, such schemes fall prey to multifarious complicated reflections, occlusions, and long-term signal dependencies. In this proposal, we seek to leverage a rich array of results in the theories of information, detection, large deviation, and mathematical programming to provide an architectural framework for sensor-based indoor location detection.
Research. We propose to investigate the fundamental performance limits of indoor location systems in terms of their robustness and resolution and to develop new SensorNet architectures that will approach these limits. Within this context, we present two cornerstone approaches to indoor location detection that we believe will frame a wide variety of hybrid practical solutions. The first solution makes use of robust identifying codes to provide a robust system that requires little knowledge of the environment’s signal space and, thus, is resilient to changes in this environment. Such a system is designed to be easily and rapidly assembled, in a self-organizing fashion, and makes use of code redundancy to tolerate environmental variabilities. Our second solution involves optimizing the SensorNet deployment in order to improve the accuracy of the location detection system, sometimes drastically so. We envision that typical applications would involve a hybrid of these two schemes, trading off resolution and accuracy for a desired level of robustness in the face of system energy and power constraints.
Broader Impact. Accurate location detection enables a host of interesting and important applications, for example: (i) smart office applications, such as providing active maps to visitors or printing to a nearby printer; (ii) tracking expensive hospital equipment; (iii) finding faulty sensors in an existing SensorNet; (iv) enabling supply chain coordination with inventory tracking; (v) assisting in victim rescue and evacuation in confined disaster areas, or assisting blind users with indoor navigation. In addition to traditional educational and dissemination goals, we plan to leverage existing contacts with our CISE center, local industry, a university NSF IGERT award, and outreach contacts (such as a SensorNet workshop being organized by one of the PIs) to facilitate the expedient dissemination of our results and corresponding impact on our educational goals. We also plan to build an open testbed for our system that will be incorporated into relevant class and research projects.